Predicting Cash Flows of Interest Payments for a 500M EUR Mortgage Portfolio
In this project on survival analysis, Michiel delved into predicting the expected cash flows from interest rate payments using a combination of statistical techniques and machine learning models, namely Cox Proportional Hazards (CoxPH), DeepSurv, and DeepHit. The CoxPH model allowed for the assessment of various factors influencing the timing of interest payments, while DeepSurv and DeepHit introduced deep learning to capture complex patterns, respectively.
This research involved extensive data analysis, from preparation to feature engineering, and model optimization. The outcome offered several opportunities to enhance decision-making in interest-rate sensitive portfolios through advanced predictive analytics.
